Today in Generative Media

Denmark's deepfake copyright law; Six-figure faceless AI; Prime video adopts genAI

[This newsletter will be taking a few days’ vacation, returning on July 8 or so.]

News and Opinion

• Denmark to tackle deepfakes by giving people copyright to their own features (The Guardian)

• The Rise Of Six-Figure Faceless AI Video Creators (Forbes)

• Prime Video Is Increasingly Using Generative AI To Enhance In-App Viewing — Here’s Where To Expect The Tech From Live Recaps To Language Dubbing (Deadline)

• AI Redefines Filmmaking Landscape, Expert Says, Unlocking Creativity and Sparking Ethical Debates (Dataconomy)

• I’ve Spent My Life Measuring Risk. AI Rings Every One of My Alarm Bells (Time)

• The Oversight Board calls Meta's uneven AI moderation 'incoherent and unjustifiable' (Engadget)

• AI’s Biggest Threat: Young People Who Can’t Think (Wall Street Journal)

• Khan Academy CEO predicts AI in the classroom will be like 5 'amazing graduate students' assisting teachers (Business Insider)

• Why Children Learn Language Faster Than AI (Neuroscience News)

• What AI Still Can’t Do — And Why It Matters More Than Ever (Forbes)

• People use AI for companionship much less than we’re led to believe (TechCrunch)

• 15 new jobs AI is creating - including 'Synthetic reality producer' (ZDNet)

Software

• Runway Game Worlds - Early Access (X)

• Flux.1 Kontext [dev] Day-0 Support in ComfyUI (comfy.org)

• OmniGen2: a powerful and efficient unified multimodal model (replicate.com)

• Try on looks and discover your style with Doppl (Google Keyword blog)

• Hollywood duo wants you to own your AI self (Axios)

• Introducing Inworld TTS: State-of-the-art and radically affordable Voice AI for every developer.

• Amazon's Ring can now use AI to 'learn the routines of your residence' (The Register)

• Build and share AI-powered apps with Claude (Anthropic)

• apps built as artifacts now have the ability to run their own prompts against Claude via a new API. (Simon Willison’s blog)

• Desktop Extensions: One-click MCP server installation for Claude Desktop (Anthropic)

Research

• BulletGen: Improving 4D Reconstruction with Bullet-Time Generation (project page)

• AnimaX: Animating the Inanimate in 3D with Joint Video-Pose Diffusion Models (project page)

• Light of Normals: Unified Feature Representation for Universal Photometric Stereo (project page)

• Stereo Any Video: Temporally Consistent Stereo Matching (project page)

• Guidance in the Frequency Domain Enables High-Fidelity Sampling at Low CFG Scales (arXiv)

• Virtual Memory for 3D Gaussian Splatting (arXiv)

• Reward models (RMs) are the moral compass of LLMs – but no one has x-rayed them at scale. We just ran the first exhaustive analysis of 10 leading RMs, and the results were...eye-opening. Wild disagreement, base-model imprint, identity-term bias, mere-exposure quirks & more (X)

Misc

• Lovely hack to pay less to OpenAI when transcribing. Just speed up the audio you send to them. (X)

• +1 for "context engineering" over "prompt engineering". (X)